This Week in AI
This Week in AI

Why Anthropic is watermarking every word Claude writes | E26

August 13, 2026

AI Summary

5 min read

Anthropic announced it will watermark every piece of text generated by its Claude models, embedding a digital signature at the model level that travels with text when copied and pasted. The move complies with the EU AI Act's transparency code, which took effect in August 2024 and requires AI companies to mark AI-generated content in ways other systems can identify. Anthropic is using the C2PA open standard and says the watermark will persist through some editing, applying to all models released after August 2nd, with older models updated retroactively.

The panel of founders on This Week in AI—Will Brick (Xa), Josh Serrota (Aragon), and Kosh Ali (TaxGPT)—discussed whether this is a net positive or just another regulatory burden. Most agreed the watermark itself is a small technical lift that doesn't meaningfully slow Claude down. But the conversation quickly widened to the deeper cultural and strategic implications.

The EU's anti-AI posture vs. opt-in pragmatism

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

What you'll learn

  • 1 Timestamped Outline
  • 2 (00:00) **Episode Open & Host Introduction** - Jason Calacanis sets up the show format and introduces the panel of builder-founders
  • 3 (03:38) **Anthropic's Claude Watermarking Announcement** - Anthropic will watermark all text from Claude models using the C2PA open standard to comply with EU AI Act transparency code
  • 4 (06:44) **Panel Reacts: EU Regulation vs. Opt-In Debate** - The guests debate whether watermarking should be mandated by regulation or left as an opt-in feature
  • 5 (10:57) **The Slippery Slope of Regulation** - Discussion of whether watermarking regulation will push users toward open-source models and create enforcement challenges
  • 6 (12:52) **Business Impact of Watermarking on Enterprise** - Kosh explains how watermarking could affect enterprise client relationships and create new pricing dynamics
  • 7 (15:20) **EU as a Data Point on AI Diffusion** - Josh argues the watermarking debate reveals more about Europe's overall AI posture than about watermarking itself

+ Full timestamped outline available in the app

Guests on this episode

Show Notes

This Week In AI is made possible by:

Paypal Open


Today’s show:

Anthropic’s new Claude watermark travels around any and all copy-and-pasted text. Is this a logical response to AI slop, or more overreach from those regulation-loving Europeans? The panel agrees that identifying AI outputs has a place, but divides on how the world should get there.

PLUS everyone’s favorite section of Mark Zuckerberg’s 6500-word techno-optimist AI manifesto, exploring the frontier labs’ blatant hypocrisy around distillation, and why Texas is hitting the pause button on 474 GW of proposed data centers.


Guests:

Will Bryk: https://x.com/WilliamBryk

Exa: https://exa.ai/

Josh Sirota: https://x.com/joshua_sirota

Eragon: https://www.eragon.ai/

Kashif “Kash” Ali: https://x.com/ChKashifAli

TaxGPT: https://www.taxgpt.com/


Relevant Links:

Anthropic: How Claude marks AI-generated content: https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content

Coalition for Content Provenance and Authenticity (C2PA): https://c2pa.org/

EU AI Act: Article 50 transparency obligations: https://artificialintelligenceact.eu/article/50/

TechCrunch: Pangram Raises $9M: https://techcrunch.com/2026/07/29/as-ai-content-floods-the-internet-pangram-raises-9m-to-detect-it/

TechCrunch: Anthropic’s landmark copyright settlement is approved: https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/

World ID iris scanning orb: https://world.org/world-id

Time: Sam Altman voices concern about “Dead Internet Theory”: https://ti

This Week in AI